Queryable, verified video

Compression vendors stop at a smaller file. The pain in a real video or PACS lake isn't 10% fewer bytes — it's that the footage is opaque, unaddressable, and unauditable.AT-1 wraps any codec's frames (H.264 / AV1 / DICOM — we never re-encode the pixels) in a container that turns a query into an action.

Find it → pull it out → prove it

The search is the easy part. The value is what you do next: extract the exact frame or clip as a standalone, independently verified artifact — without ever decoding or rehydrating the rest of the file.

# 1. wrap a video's frames in a queryable, verified container (no pixel re-encode)
at1 media build camera.mp4 camera.at1vid

# 2. FIND the moment — query the index, never decode the whole file
at1 media query camera.at1vid motion 1.2        # frames with activity
at1 media query camera.at1vid time 30 45        # everything 30–45s
at1 media query camera.at1vid cuts              # scene boundaries

# 3. PULL IT OUT — extract just those frames into a NEW verified clip,
#    WITHOUT rehydrating the 64 GB source (true range-GET extraction)
at1 media clip camera.at1vid event.at1vid 21,22,23,24,25
at1 media get  camera.at1vid 42 frame42.png     # one frame, reading only its bytes
at1 media export event.at1vid event.gif 0,1,2,3,4

# 4. PROVE IT — byte-exact, and a single altered byte is detected AND located
at1 media verify event.at1vid
at1 media redact camera.at1vid redacted.at1vid 0,1   # blank PII, re-stamp chain-of-custody
  • Addressability — every frame carries a byte offset, so get / clip read only the frames you asked for. Pull one event out of a ten-hour file without touching the other ten hours.
  • Query-in-place — each frame embeds a perceptual feature, so time, motion/activity, scene-cut, and search-by-image run over the index with no payload decode.
  • Extract = a new verified container — clip copies the matching frames into a fresh artifact that verifies on its own. The result of a search is a thing you can ship, store, and trust — not a list of timestamps.
  • Chain-of-custody — per-frame + global SHA-256. verify proves byte-exactness, and a single tampered byte is both detected and pinpointed to the exact frame. redact blanks PII and re-stamps the chain.

Honest scope: we carry the standard codec's payload — we do not compete with AV1/H.265 on pixel ratio. The win is the queryable + verified + addressable layer they omit: medical PACS archives, surveillance / body-cam chain-of-custody, and media asset management. Same SDK in Python (import at1_media) and the at1 media CLI.